Looking for a quick way to find odd / even numbers

[Johnny Starr]

I'm working on a routine for an Atari 2600 game. The portion of the code is the 192 visible scanlines. I'm attempting to achieve a visual effect by updating the background every other scanline.

The scanline loop simply counts from 192 to 0. In this scheme, #192 represents the top line, #0 the bottom or last line. It seems that there should be a way to determine odd / even numbers to accomplish this. I'm not sure if this is considered "parity" or not, but I haven't found the right flag for odds / evens.

[Aslak3]

Johnny Starr wrote:

I'm working on a routine for an Atari 2600 game. The portion of the code is the 192 visible scanlines. I'm attempting to achieve a visual effect by updating the background every other scanline.

The scanline loop simply counts from 192 to 0. In this scheme, #192 represents the top line, #0 the bottom or last line. It seems that there should be a way to determine odd / even numbers to accomplish this. I'm not sure if this is considered "parity" or not, but I haven't found the right flag for odds / evens.

Parity is a count of the number of set bits. For oddness/eveness, just test the lowest bit. If it is 0, it's even. if it is 1 it is odd.

Edit: But from the sounds of it you just want to update all the even lines (say) then the odd ones. It is so simple I must be misunderstanding, but can't you simply add 2 on each loop iteration?

[BigDumbDinosaur]

Johnny Starr wrote:

I'm working on a routine for an Atari 2600 game. The portion of the code is the 192 visible scanlines. I'm attempting to achieve a visual effect by updating the background every other scanline.

The scanline loop simply counts from 192 to 0. In this scheme, #192 represents the top line, #0 the bottom or last line. It seems that there should be a way to determine odd / even numbers to accomplish this. I'm not sure if this is considered "parity" or not, but I haven't found the right flag for odds / evens.

It's very simple:

Code: Select all

         lda scanline          ;current scan line number
         lsr a                 ;shift bit 0 to carry
         bcc is_even           ;even number
         bcs is_odd            ;odd number

Alternatively, with the 65C02/65C816:

Code: Select all

         lda scanline          ;current scan line number
         bit #%00000001        ;is bit 0 set?
         beq is_even           ;no, even number
         bne is_odd            ;yes, odd number

[Johnny Starr]

Of course! Shift right, check carry. That's what I get for overthinking it huh?

[GARTHWILSON]

or AND#1, BNE/BEQ. The AND# takes 2 clocks, and if the conditional branch is not taken, it's 2 more, or if the branch is taken, it's usually 3 more instead of 2. You can probably get away with one branch instruction, as the drop-thru is automatically the other condition.

[BigDumbDinosaur]

GARTHWILSON wrote:

or AND#1, BNE/BEQ. The AND# takes 2 clocks, and if the conditional branch is not taken, it's 2 more, or if the branch is taken, it's usually 3 more instead of 2. You can probably get away with one branch instruction, as the drop-thru is automatically the other condition.

BIT immediate is the least destructive and produces the same result.

[BigEd]

But you can't do that on a proper 6502!

[teamtempest]

If all you're interested in is odd/even and not the particular scanline, you can do this on all 65xx-family processors:

Code: Select all

  LDA #%00000001
  BIT  scanline
  BEQ doeven
  BNE do odd

It's not as fast as load-and-shift, though.

BTW, 0..192 is 193 lines, not 192. Just sayin'.

[BigDumbDinosaur]

BigEd wrote:

But you can't do that on a proper 6502!

Right. I was thinking 'C02 or '816. BIT immediate is one of the many reasons to not use the NMOS 6502 if you have a choice.

[White Flame]

If you store the value 1 to a zero page memory location const1, then you can do "BIT const1" to perform the test without modifying the accumulator on ye olde 6502.

[barrym95838]

I was thinking the same thing, White Flame. Or just find a 1 somewhere in your program or the firmware ... there are bound to be several. Code-as-constants is a tried-and-true tradition in the 65xx world, even more so than self-modifying.

Mike